Early Careers hiring looks very different to how it did three years ago.
None of that will be news if you're hiring at this level. Volumes are up. AI tools have made it much easier to research, structure and refine an application or prepare for a standard interview question, so the differences between candidates can be harder to read from the application alone.
What gets less attention is that the process of sorting them has barely changed. Most firms are still running a version of the funnel they built before any of this happened.
If you've wondered whether yours is still doing its job, so have your peers. Financial institutions in our 2026 research put their own confidence at 55 out of 100 when asked how well their assessment predicts on-the-job performance. Candidates are even less confident: only 7% think the current assessment process measures their ability to do the job very accurately.
A process can run exactly as designed and still measure the wrong thing. Here are seven ways to improve yours.
1. Screen last, assess first
Which assessment methods give you the clearest indication of how someone will actually perform?
In our 2026 employer polling, in-person interviews and role-specific simulations were identified most often as the strongest indicators of future performance. CVs, aptitude tests and pre-recorded interviews received far fewer selections.
Now think about where those sit in your own process. CVs, tests and automated stages often sit much earlier, where some of the biggest cuts can happen. Most processes make their biggest cut at their weakest stage.
That caps everything after it. However good your assessment centre is, it only ever sees the people who got through the first filter, and no amount of rigour at later stages recovers someone you lost at stage one. You won't know you lost them, either. Rejected candidates don't show up in performance data.
So: map which stage removes the most people, then ask what evidence that stage actually runs on. Where those two don't match, that's where to start.
2. Count what your biggest cut is costing you
Whichever stage removes the most people is the one carrying your process. If it isn't sorting well, everything downstream starts from a weaker pool than you think it does.
You still pay full price for it. Interview slots, assessment centre places and hours from hiring managers. None of that shows up as a problem in your numbers, because nothing about the stage looks broken. It carries on producing a shortlist of exactly the size you asked for, on time, every year.
Take one intake and work backwards. How many of your strongest hires would have got through that stage if you'd run them through it blind?
3. Check whether you're measuring ability or access
Some of what your assessment rewards is ability. Some of it is preparation that not everyone has access to.
Our research suggests candidates feel this keenly. Students name preparation rather than potential as one of the barriers they keep hitting, and say employers put too much weight on interview confidence and "fit". One of the students stated:
"Most 'high performing' candidates just memorised stock answers to questions that bankers want to hear. It feels like banks just want to recruit robots."
A polished answer can reflect ability, but it can also reflect preparation and access to support. That makes it worth asking how much success at each stage depends on knowing what to expect beforehand.
Go through your stages and ask what a candidate would need to do well in each one. If coaching, contacts or prior exposure to the process materially improve someone's chances, that's worth understanding alongside the capability you're trying to measure.
4. Assess on evidence, not impressions
How a candidate comes across matters. Communication is a real part of most jobs, and an interview is the right place to test it.
The trouble starts when that judgement gets asked to carry more than it can. Presence and articulacy don't necessarily tell you how someone handles information that changes halfway through, or what they do when their first approach doesn't work. Most assessment stages end in a judgement, and very few end in a record of what the candidate actually did.
Our research suggests employers are already pulling those apart, describing a shift toward observable performance, communication and decision-making rather than polished or rehearsed answers.
The practical difference comes later. A record gives you tangible proof of who performed and who didn't, and something to point at when a hiring manager asks why one person went through and another didn't. An impression gives you someone's memory of a conversation that happened six weeks ago.
Take a recent decision and ask what it rested on. If the evidence for someone's judgement turns out to be evidence of how well they spoke, you're scoring two different things as one.
5. Test for the role, not the funnel
A trading seat, an M&A desk and a quant team don't need the same person, yet parts of the assessment process can still look remarkably similar across them. One set of tests, one interview format, one assessment centre, run across divisions that want genuinely different things.
That isn't laziness. Running separate assessments per desk costs more and scales worse, and at these volumes that's a real constraint rather than an excuse.
It does mean you end up measuring the average of jobs that aren't alike. The skills students tell us get overlooked most often are practical problem solving, decision-making under pressure and commercial judgement, and those three look nothing like each other on a trading floor and in an advisory team.
Ask a desk head what separates their best junior from their average one. Then work out where in your process that difference would ever have shown up.
6. Watch what candidates do with feedback
At this level you're hiring for how fast someone learns, not for what they already know.
Learning agility is the skill students most often say gets missed, named by 45% of them, and it's also the hardest thing on that list to assess. A single sitting can't show change over time. Asking someone how they respond to feedback only tells you how well they can describe responding to feedback.
Watching is different. Give a candidate specific feedback after a practical exercise, let them have another go, and the gap between the two attempts is the evidence.
"The second attempt is usually the telling one. You can have two candidates score identically the first time round, then you give them the same feedback and watch what happens. One of them applies it straight away, the other doesn't. That tells you far more about how they'll develop on the desk than either score does."
Anthony Cheung, Chief Content and Culture Officer, AmplifyME
Yes, it adds a step to a process you're probably trying to shorten. But even a short feedback-and-retry stage can show you something a single score can't: whether the candidate can take new information and improve what they do next.
Build one stage where feedback is given and then acted on, even a short one. What someone does with it is a signal you won't get anywhere else.
7. Compare against everyone, not just the room
When assessment centres score people against whoever else turned up that day, it makes sense when you're filling a set number of places and need a ranking by five o'clock.
The trouble is that the room varies. A strong candidate in an unusually strong group can miss out. A weaker one on a quiet day gets through. Neither looks like a mistake at the time, because the score sheet reads the same either way.
A fixed benchmark takes the day out of it. Measure everyone against the same reference population and a strong candidate is strong regardless of who else was sitting there.
That becomes particularly valuable as you build a larger performance dataset over time. Instead of only comparing candidates with the people in front of you on the day, consistent assessment can give you a much broader reference point for what performance looks like.
One global investment bank featured in our Annual Report has now used simulation-based assessment with more than 250,000 candidates globally over several years. The programme has also seen a 30% increase in applications from non-traditional backgrounds.
Check whether your scoring is relative or absolute. If two candidates performing identically on different days wouldn't get the same outcome, the day is part of your decision.
The application has changed. Assessment mostly hasn't.
None of this means rebuilding a process from scratch. Reorder a stage. Send real feedback instead of a template. Put one live exercise where a written one used to be.
Small changes can give you a different view of the candidates in front of you, and help you understand whether the stages doing the most filtering are also giving you the information you trust most.
The candidates you hire aren't the only measure of how well your process is working. The people it filters out matter too.
Our 2026 Annual Report, The Future of Early Careers Assessment in Finance, draws on 750 students and recent graduates and more than 60 employer and industry perspectives. This piece is about what to check in your own process. The report covers what firms are doing about it, including the five priorities shaping Early Careers assessment.